Odd return value in snowflake query

[PHilden-RP]

Running the below query on a snowflake database returns the NULL values of grp, where I would expect it would return the max val for which grp is not NULL

tmp <-
  tibble(
    grp = c(NA, "grp1", "grp2", "grp2"),
    val = c(4, 3, 2, 1)
  ) %>% 
  mutate(id = "abc") %>% 
  relocate(id)

tmp_con <- copy_to(con, tmp, I("DATABASE.SCHEMA.TMP"), overwrite = TRUE)

tmp_con %>% 
  group_by(id) %>%
  slice_max(val[!is.na(grp)], with_ties = FALSE) %>% 
  ungroup()

Currently results from the above query

I would have expected this to return the 2nd row in tmp, which is the one with the largest val in which grp is not NA